Pytorch implementation of the Neural Overlapping Community Detection method from "Overlapping Community Detection with Graph Neural Networks".

The main algorithm and other utilities are implemented in the nocd package that can be installed as

python setup.py install

A Jupyter notebook interactive.ipynb contains the code for training the model and analyzing the results.

Experiments in the paper have been performed using an older TensorFlow version of the code that can be found here.

numpy=1.16.4
pytorch=1.2.0
scipy=1.3.1

Please cite our paper if you use the code or the datasets in your own work

@article{
    shchur2019overlapping,
    title={Overlapping Community Detection with Graph Neural Networks},
    author={Oleksandr Shchur and Stephan G\"{u}nnemann},
    journal={Deep Learning on Graphs Workshop, KDD},
    year={2019},
}